The Personal Training Report 2026
Routines are designed for four or five days a week. The real median is two. That distance between what gets prescribed and what gets executed organizes everything below.
This report measures how training gets programmed and how it gets done, not how many people use Kaizer. These are aggregate, anonymized distributions, and every figure travels with the sample it came from.
This report analyzes a sample of 10,529 routines, 701,376 prescribed exercises and 1,928,481 programmed sets inside Kaizer between 30 July 2025 and 5 August 2026. These are aggregate, anonymized data from a sample of the platform, not from all of it. No individual trainer or client is identifiable.
Data window: 30 July 2025 to 5 August 2026 · Cut-off: 5 August 2026
Key terms
- Prescription
- One exercise assigned inside a session: the exercise, its sets and its reps exactly as they were saved. It is the unit that measures what gets programmed.
- Client-week
- One calendar week of one specific client in which they logged at least one session. It is the unit that measures what gets trained.
- Median
- The value that splits the sample into two equal halves. Unlike the average, a single extreme case does not move it.
Coaches program twice what clients train
79.2% of routines are designed for three to five training days a week. Five days is the most common choice at 36.1%; four days sits at 26.1% and three days at 17.1%. It is ambitious programming and perfectly orthodox.
On the other side of the same product, the median number of days trained per week is two. Only 42.1% of weeks with activity reach three days or more, and 24.0% reach four or more.
The two figures do not come from the same base — one counts routines, the other counts client-weeks — so subtracting them does not produce a compliance rate and nobody should use them that way. The plan and the behaviour live on different scales.
Real frequency is computed only over weeks in which the client logged at least one session. Weeks at zero are excluded. So the median of two overstates consistency rather than understating it.
What gets planned and what gets logged
Programmed, per routine
n = 10,477 routines · Jul 2025 to Aug 2026
Trained, per week with activity
n = 8,073 active client-weeks · Jul 2025 to Aug 2026
Two series with different bases, drawn separately on purpose. Among routines: five days 36.1%, four days 26.1%, three days 17.1%. Among client-weeks with activity: 42.1% reached three days or more and 24.0% reached four or more, with a median of two days. Bars in each panel are scaled to that panel's own highest value.
The bench press is the seventh most programmed exercise
Across 589,497 prescriptions from the official catalogue, the five most programmed exercises are the cable triceps extension (3.36%), the lat pulldown (3.09%), the lateral raise (2.82%), the machine chest fly (2.82%) and the incline dumbbell press (2.8%).
The bench press comes seventh, at 2.65%. The back squat is twenty-second, at 1.19%.
Real programming does not look like the poster of the big five. Personal training is being written on isolation work and machines, and it treats the big lifts as one more piece of the plan rather than its backbone.
The most prescribed exercises
n = 589,497 prescriptions from the official catalogue · Jul 2025 to Aug 2026
Share of official-catalogue prescriptions taken by each exercise: cable triceps extension 3.36%, lat pulldown 3.09%, lateral raise 2.82%, machine chest fly 2.82%, incline dumbbell press 2.8%, bench press 2.65% (seventh in the overall ranking) and back squat 1.19% (twenty-second). Bars are scaled to the highest value in the list.
The dumbbell beat the barbell
Across 629,223 equipment assignments in prescribed exercises, the dumbbell leads with 24.1%. Machines (19.7%) and cables (19.6%) follow. The barbell comes fourth at 14.2%, and bodyweight fifth at 10.4%.
Added together, machines and cables are 39.3% of all programming: almost four out of ten prescriptions happen on guided equipment.
For a gym this is a purchasing decision: the free-weight floor and the machine floor are not receiving the same amount of programming.
How equipment is distributed
n = 629,223 equipment assignments · Jul 2025 to Aug 2026
Share of equipment assignments: dumbbell 24.1%, machine 19.7%, cable 19.6%, barbell 14.2% and bodyweight 10.4%. The remaining 12.0% is spread across smaller categories that are not in the chart. Bars are scaled to the highest value.
One in seven prescribed exercises never gets done
Across 119,172 exercises actually executed, 84.7% are completed, 1.6% end up partial and 13.8% are marked as skipped.
The skip ranking is coherent: cable crossover (24.4%), treadmill (20.3%), plank (19.9%), standing calf raise (19.5%) and dips (19.2%).
What gets abandoned is the accessory work and everything without external load: cardio, core and bodyweight. If a plan needs something to survive a rushed session, that something had better not sit at the end.
The most skipped exercises
n = 119,172 executed exercises · only exercises with 300 or more records
Skip rate of each exercise over its own logged executions: cable crossover 24.4%, treadmill 20.3%, plank 19.9%, standing calf raise 19.5% and dips 19.2%. Only exercises with 300 or more executions are included, so the catalogue's long tail cannot invent a ranking. Bars are scaled to the highest value.
Personal training happens at 7 p.m. on Monday
Across 13,839 sessions from clients with their own declared time zone, Monday and Tuesday concentrate 41.0% of all sessions — 20.5% each. The entire weekend adds up to just 13.5%.
The 6 to 9 p.m. band concentrates 32.7% of sessions, and the peak hour is 7 p.m., with 9.7%.
The training week starts strong and fades. A plan that spreads volume evenly from Monday to Sunday is fighting the way people actually train.
The shape of the week
n = 13,839 sessions from clients with their own declared time zone
Share of logged sessions by part of the week: Monday and Tuesday 41.0%, Saturday and Sunday 13.5%, and Wednesday to Friday 45.5%, which is the difference between the two measured blocks and the total. Clients whose time zone reads as the default UTC are excluded, because there the local hour is not data but an unset field. Bars are scaled to the highest value.
AI-assisted programming quadrupled in nine months
In November 2025, 5.7% of new routines were generated with AI. In August 2026, 22.5%.
The series starts in November on purpose. Before that date the product mix was different: monthly routine volume multiplied by nine between September and December 2025, when coach-led usage came in. Chaining the two periods would produce a drop in adoption that never happened.
One in five new routines is now born with AI assistance. It is still a minority. Adoption grows fast without becoming the default.
New routines generated with AI
n = 10,529 routines · monthly series from November 2025
Share of new routines generated with AI: 5.7% in November 2025 and 22.5% in August 2026. The series excludes the months before November 2025 because the product had a different usage mix and the comparison would be false. Bars are scaled to the highest value.
The baseline: three sets of six to twelve reps
Across 1,928,481 programmed sets — the largest sample in this report — 87.1% fall between 6 and 12 reps: 59.5% between 6 and 8, and 27.7% between 9 and 12. The median is 8. Pure strength work, 1 to 5 reps, is only 5.4%.
Across 691,669 prescribed exercises, three sets is the undisputed standard at 72.5%. Four sets appears in 11.6% and two sets in 6.2%.
It is the least surprising figure in the report and the one with the largest sample. When someone argues about rep ranges, this is what actually gets written into routines.
Reps per programmed set
n = 1,928,481 programmed sets · Jul 2025 to Aug 2026
Share of programmed sets by rep range: 6 to 8 is 59.5%, 9 to 12 is 27.7%, 13 or more is 7.4% and 1 to 5 is 5.4%. The 6-to-12 band holds 87.1%. Percentages are rounded, so a sum of categories can drift by a tenth. Bars are scaled to the highest value.
Other signals from the sample
- The four-week mesocycle dominates: 46.4% of routines run exactly four weeks, and eight weeks is the runner-up at 17.6%.
- n = 10,529 routines
- The session is designed at 60 minutes: 42.1% of generator requests ask for one hour. Then come 75 minutes (18.6%), 90 minutes (13.8%) and 45 minutes (13.2%).
- n = 1,831 routine requests · the field mixes minutes and seconds and was normalized to minutes
- 71.1% of prescriptions never received a manual edit. The trainer edited 8.7% and the client themselves edited 20.2%.
- n = 701,376 prescriptions
- 10.0% of prescriptions use an exercise created by the trainer, outside the official catalogue.
- n = 701,376 prescriptions
How to read this data
What this report shows
Distributions and patterns: percentages, medians and relative rankings inside the analyzed sample. We do not publish absolute counts of users, routines or sessions derived from this sample, because the sample is not the population.
Where the data comes from
From Kaizer's product database, over the period 30 July 2025 to 5 August 2026. Kaizer is migrating its data infrastructure and this analysis runs on the already-migrated subset. That subset is large and sufficient for measuring distributions — almost two million programmed sets — but it is a portion of the platform's total. That is why every figure is expressed as a proportion of its own sample, and every section declares the N that belongs to it.
What is not here
There is no geographic distribution and no roster size per trainer: the migrated sample does not represent them reliably, and we would rather not publish them than publish them badly. There is also no adherence rate over assigned sessions, because a session is only recorded once completed and that ratio would mean nothing.
Anonymity
No figure in this report allows identifying a trainer or a client. Every query is an aggregation, and no category is published with fewer than 150 observations.
Rounding
Percentages are rounded to one or two decimals depending on the case, so a sum of categories can drift a tenth from the total. Where a category was obtained by difference, the chart's text equivalent says so.
Kaizer is the platform where these routines were written: programming, tracking and client management in one place.
